temporalio.contrib.google_adk_agents.TemporalModel

class documentation

class TemporalModel(BaseLlm): (source)

Constructor: TemporalModel(model_name, activity_config, summary_fn, streaming_topic, ...)

A Temporal-based LLM model that executes model invocations as activities.

Method	`__init__`	Initialize the TemporalModel.
Async Method	`generate_content_async`	Generate content asynchronously by executing model invocation as a Temporal activity.
Instance Variable	`_activity_config`	Undocumented
Instance Variable	`_model_name`	Undocumented
Instance Variable	`_streaming_batch_interval`	Undocumented
Instance Variable	`_streaming_topic`	Undocumented
Instance Variable	`_summary_fn`	Undocumented

def __init__(self, model_name: str, activity_config: ActivityConfig | None = None, *, summary_fn: Callable[[LlmRequest], str | None] | None = None, streaming_topic: str | None = None, streaming_batch_interval: timedelta = timedelta(milliseconds=100)): (source) ¶

Initialize the TemporalModel.

Streaming is selected by the caller via the ADK generate_content_async(stream=True) argument; no plugin-level flag is needed.

Parameters
model_name:`str`	The name of the model to use.
activity_config:`ActivityConfig \| None`	Configuration options for the activity execution.
summary_fn:`Callable[[LlmRequest], str \| None] \| None`	Optional callable that receives the LlmRequest and returns a summary string (or None) for the activity. Must be deterministic as it is called during workflow execution. If the callable raises, the exception will propagate and fail the workflow task.
streaming_topic:`str \| None`	Stream topic to publish raw `LlmResponse` chunks to when streaming. Required when callers invoke `generate_content_async(stream=True)`; if `None`, the streaming call raises before scheduling an activity. The workflow must host a `temporalio.contrib.workflow_streams.WorkflowStream` to receive the publishes; otherwise the signals are unhandled and dropped. Streaming support is experimental and may change in future versions.
streaming_batch_interval:`timedelta`	Interval between automatic flushes for the stream publisher used by the streaming activity. Streaming support is experimental and may change in future versions.

Raises
`ValueError`	If both `ActivityConfig["summary"]` and `summary_fn` are set.

async def generate_content_async(self, llm_request: LlmRequest, stream: bool = False) -> AsyncGenerator[LlmResponse, None]: (source) ¶

Generate content asynchronously by executing model invocation as a Temporal activity.

Parameters
llm_request:`LlmRequest`	The LLM request containing model parameters and content.
stream:`bool`	Whether to use the streaming activity. When `True`, each chunk is also published to `streaming_topic` (if set) for external consumers. Streaming support is experimental and may change in future versions.

Returns
`AsyncGenerator[LlmResponse, None]`	Undocumented

Yields
The responses from the model.

_activity_config = (source) ¶

Undocumented

_model_name = (source) ¶

Undocumented

_streaming_batch_interval = (source) ¶

Undocumented

_streaming_topic = (source) ¶

Undocumented

_summary_fn = (source) ¶

Undocumented